Ближайшие семинары
Календарь семинаров
Список семинаров
Архив по годам
Регистрация семинара

Ближайшие семинары

Для просмотра файлов Вам могут потребоваться

Семинар лаборатории ПреМоЛаб
12 февраля 2014 г. 17:00, г. Москва, Институт проблем передачи информации им. А. А. Харкевича РАН (Б. Каретный пер., 19, метро «Цветной бульвар»), ауд. 615

Accelerated, Parallel and Proximal Coordinate Descent

Peter Richtarik

School of Mathematics University of Edinburgh

Количество просмотров:
Эта страница:173
Youtube Video:

Аннотация: We propose a new stochastic coordinate descent method for minimizing the sum of convex functions each of which depends on a small number of coordinates only. Our method (APPROX) is simultaneously Accelerated, Parallel and PROXimal; this is the first time such a method is proposed. In the special case when the number of processors is equal to the number of coordinates, the method converges at the rate $2\bar{\omega}\bar{L} R^2/(k+1)^2 $, where $k$ is the iteration counter, $\bar{\omega}$ is an average degree of separability of the loss function, $\bar{L}$ is the average of Lipschitz constants associated with the coordinates and individual functions in the sum, and $R$ is the distance of the initial point from the minimizer. We show that the method can be implemented without the need to perform full-dimensional vector operations, which is considered to be the major bottleneck of accelerated coordinate descent. The fact that the method depends on the average degree of separability, and not on the maximum degree of separability, can be attributed to the use of new safe large stepsizes, leading to improved expected separable overapproximation (ESO). These are of independent interest and can be utilized in all existing parallel stochastic coordinate descent algorithms based on the concept of ESO.
Presentation is available here:

Язык доклада: английский

ОТПРАВИТЬ: FaceBook Twitter Livejournal
Обратная связь:
 Пользовательское соглашение  Регистрация  Логотипы © Математический институт им. В. А. Стеклова РАН, 2019